Statistical results used to determine model selection for mixed-effects models. Full models included surgery group (control and ovariectomized), recording condition (dyad, elicited, and isolation), and their interaction as fixed effects and individual rat as the random effect. Reduced models for principal frequency, maximum frequency, frequency bandwidth, frequency standard deviation, sinuosity, and mean only included recording condition as the fixed effect and individual rat as the random effect. Reduced model for slope included both recording condition and surgery group as fixed effects and individual rat as the random effect. ANOVAs were used to test full versus reduced models. No significance resulted in selection of reduced models. Numbers were rounded to the third decimal place.

ANOVAs of full vs reduced models for acoustic variables with nonsignificant interaction terms.
USV Acoustic parameter	npar	AIC	BIC	logLik	deviance	Chisq	Df	Pr(>Chisq)
Principal Frequency
	5	135623.862	135662.977	−67806.931	135613.862
	8	135625.190	135687.773	−67804.595	135609.190	4.672	3	0.197
Maximum Frequency
	5	142966.900	143006.015	−71478.450	142956.900
	8	142967.147	143029.730	−71475.573	142951.147	5.753	3	0.124
Bandwidth Frequency
	5	136824.889	136864.004	−68407.445	136814.889
	8	136823.509	136886.093	−68403.755	136807.509	7.380	3	0.061
Slope
	6	264076.941	264123.879	−132032.471	264064.941
	8	264078.185	264140.768	−132031.093	264062.185	2.756	2	0.252
Frequency Standard Deviation
	5	93005.222	93044.337	−46497.611	92995.222
	8	93006.229	93068.812	−46495.114	92990.229	4.993	3	0.172
Sinuosity
	5	41772.678	41811.793	−20881.339	41762.678
	8	41769.172	41831.755	−20876.586	41753.172	9.507	3	0.023*
Mean Power
	5	120865.701	120904.816	−60427.851	120855.701
	8	120866.006	120928.589	−60425.003	120850.006	5.696	3	0.127

^*p <.05